arXiv cs.AI· Ziwei Wu, Jingrui He·· 10 小时前AI 评分26
FAIDER：面向大语言模型推荐器的公平性感知条目级解码方法
FAIDER: Fairness-Aware Item-Level Decoding for Large Language Model Recommenders
AI 导读
FAIDER 通过选择完整条目候选而非操纵 token 级 logits 或反复修复输出，将公平性融入推荐列表构建。该方法用 conformal 启发的经验校准确定跨群体语义差异阈值，在线解码时采用硬过滤或经对偶启发反馈调整的软惩罚，以逼近预设违规率。在多个推荐基准和 LLM 主干上，Soft FAIDER 的语义公平性与排序质量均优于所评估的 prompt 修复基线。
正文 · 原文
Abstract:Large language model (LLM) recommenders can produce semantically different recommendations when a user's protected attribute changes, even if the interaction history remains fixed. Controlling this disparity is challenging when model weights and full-vocabulary logits are inaccessible. We propose FAIDER (Fairness-Aware Item-level DEcoding for Recommendation), which incorporates fairness into list construction by selecting complete item candidates rather than manipulating token-level logits or repeatedly repairing finished outputs. The central idea is to separate the operational criterion for a violation from how strongly generation is constrained. Conformal-inspired empirical calibration fixes a cross-group semantic-disparity threshold, while online decoding applies either hard filtering or a soft penalty adjusted through dual-inspired feedback to target a prescribed violation rate. Separately, we establish a pathwise guarantee: under a uniform enforceability assumption, the soft decoder's empirical violation rate exceeds its target by at most $O(1/T)$. Experiments across recommendation benchmarks and LLM backbones show that Soft FAIDER improves both measured semantic fairness and ranking quality over the evaluated prompt-repair baselines. It also generally achieves a more favorable fairness-utility balance than the hard variant, indicating that stricter per-recommendation constraints do not necessarily yield lower measured disparity.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.10285 [cs.AI] |
| (or arXiv:2611.10285v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2611.10285 arXiv-issued DOI via DataCite |
Submission history
From: Ziwei Wu [view email]
[v1]
Fri, 6 Nov 2026 22:31:29 UTC (230 KB)
[v2]
Tue, 10 Nov 2026 11:21:43 UTC (232 KB)
来源：arXiv cs.AI · arxiv.org